A.I. & Optimization

Advanced Machine Learning, Data Mining, and Online Advertising Services

You May Have Been Wasting your Money on Google AdWords

In this article we discuss the effectivity of Search Engine Marketing.

To Bid or Not To Bid?

A recent study by researchers from Berkley University and eBay company analyzed the effectiveness of search engine marketing (SEM) campaigns [1]. The researchers have found that "brand keywords" advertising in which marketers bid for keywords containing their brand names like "ebay" doesn't produce measurable short-term value for the company. Researchers did a few controlled experiments in which they shut off bidding for brand keywords like "ebay shoes" on their Microsoft and Yahoo! SEM campaigns while continuing to pay for these keywords on Google AdWords. Their results show that bidding for brand keyword doesn't make significant difference on their sale rates. However, it clearly increases their costs.

One explanation is that when a user enters a query like "ebay shoes", this shows that she already knows the brand and her intention is to go to ebay website anyway. In this situation bidding for such terms and showing paid ad on the top of search results doesn't change the consumer's behavior. We should note that even if ebay bids for those types of keywords, their unpaid search results (organic results) appear right after the paid ones. The experimental results in the research paper also shows that shutting off paid search ads for brand keywords closed the costly path to the company's website but diverted traffic to natural search, which is free. On August 16, I ran a test query for "ebay" keyword a few times and found out most of time there is no ad for ebay website and the top link is the organic link to their website. See the image below which shows the results of my search query for "ebay" keyword on Google:

ebay search result
Figure 1 - Search results for "ebay" keyword.

On the other hand, searching for "amazon" often shows a paid link to amazon's website as shown in the following figure:

amazon search result
Figure 2 - Results of searching for "amazon" keyword.

On Aug 16, I searched for known brand names including "amazon", "ford", "samsung", "bmw", "toyota" and two Canadian companies: "hootsuite" and "payfirma". All searches resulted in paid ads shown at the top of the search results page directly above natural unpaid links to the companies' websites. This clearly shows that lots of marketers tend to bid and spend money for brand keywords.

Prisoner's Dilemma

As the conclusion, the researchers suggested that the known brands like "ebay" are better off to not bid for brand keywords and I assume that the result of this work changed ebay's SEM strategy. I searched for "ebay camera" on Google. Interestingly, I found that a link to "amazon" website appears on the top of the Google page results. This is because amazon has bid for "camera" keyword. See image below:

competition between amazon and ebay
Figure 3 - Competition results between amazon and ebay while searching for "ebay camera" keyword.

Although ebay doesn't bid for "ebay camera" keyword, its competitor Amazon bids apparaently for "camera" keyword and steals the traffic. This behavior can be modeled by the Prisoner's dilemma. However, to me it's not clear how much this behavior can benefit Amazon. It's important to note that in this situation the user is aware of the "ebay" brand name as she explicitly searched for "ebay camera". On the other hand, it's clear that the ad platforms benefits from this rent-seeking game.

Another example is found by my colleague Amir Hajizadeh who searched for "panago" keyword (i.e. pizza store). Image below shows the search results he got back. We see that the first result on the search page is an "AD" for another Pizza store called "Paragon". So, it's clear that this company is bidding for its competitor's brand name panago.

competition between Paragon and Panago
Figure 4 - Competition results between Paragon and Panago pizza stores while searching for "panago".


  1. Tom Blake, Steven Tadelis, and Chris Nosko, Consumer Heterogeneity and Paid Search Effectiveness: A Large Scale Field Experiment, NBER, May 2014